National scale is mainly used for low-resolution data, such as remote sensing images with the resolution of 1km or 250m, the
data sources are NOAAAVHRR, SPOT VGT, MODIS, etc. This scale of regionalization is generally carried out by
considering national meteorological elements and terrain information.
Regional scale is mainly used for high-resolution data, such as 30m, 25m, 20m and other resolution remote sensing images, the
data sources are LANDSAT TMETM+, SPOT5, CERS and 30m resolution multi-spectral images of environment or disaster
mitigation satellites. This scale of regionalization generally carried out according to vegetation, landform, soil and other
information.
Local scale is mainly used for even higher resolution data, such as 10m, 2.5m resolution remote sensing images. The main data
sources are SPOT5. In this paper, the original intention is to help to do color
balancing for Nationwide Remote Sensing imagery, to accomplish quick mosaicking of wide range remote sensing
imagery. The data resource of our test is HJ-1 with a resolution of 30m, and the color balancing algorithm is improved gain
compensation which will get better result for larger amount of data. Thus, the national scale Eco-geographic Regionalization
will be more suitable for our requirement.
2.1.2. Regionalization Principles
The definition of Eco-geographical regionalization involves a large number of standards. Regionalization in this paper is
based on the following three principles. 1 Ensure the consistency of landscape features. The elevation,
weather, soil type, hydrological conditions and other ecological characteristics of change should be fully considered. These
factors will not only affect the distribution of the natural landscape, but also the color rule of remote sensing images. If
all the elements of each partition unit are excellent consistent, color of its images will be excellent proportioned.
2 Ensure the appropriateness of the scale. To ensure that after partition, the every sub-region can be covered within a certain
number of images, the regionalization need to be national scale, to ensure the integrity of the partition without crushing area.
3 Consider DEM data and air temperature, humidity and other weather information as a leading factor.
2.1.3. Methods of Regionalization
In this paper, overlay method, leading mark method, geographic correlation analysis method and combined analysis method are
applied in combination with each other. Overlay method: Distribution maps with natural elements and
regionalization map are stacked together, and then the most overlapping lines as the basis for division will be selected.
Overlay method can reduce the subjectivity and arbitrariness, and helps to find relation between different natural phenomena.
Leading mark method: According to comprehensive analysis, the leading mark will be chosen to determine the boundaries or
part of the boundaries. Geographical correlation analysis method: Firstly compare
distribution maps with natural elements and regionalization map, then the discipline of nature regional differentiation and
interdependent relationship will be analysed, based on which the boundaries will be determined
Combined analysis: In order to classify and summarize the natural geographic areas or phenomena, tiny land category will
be removed. Gradually, the same land cover types or geomorphological units are merged together, and eco-
geographic regionalization is formed.
2.2. Nationwide Eco-geographic Regionalization
In this paper, meteorological data and DEM data are used for division with the major workflows as follows:
Second level date-value dataset of
surface climate
DEM data mean annual
temperature mean relative
humidity regionalization
boundaries dominated by climate factors
nationwide Eco-geographical regionalization
1. Overlay method 2. Leading mark method
3. geographic correlation analysis method
1. Overlay method 2. Leading mark method
3. geographic correlation analysis method
4. combined analysis method
First level
Figure 1. Process of nationwide Eco-geographic regionalization
2.2.1. Data Sources
This paper uses the date-value dataset of surface climate, the National SRTM250m DEM data.
2.2.2. Pre-processing
Firstly, we extract temperature and humidity data from 2009 Chinas date-value dataset of surface climate to calculate the
mean annual temperature and mean relative humidity, shown in Figure 2, and introduce them into ArcMap to conduct the
following reprocess:
Figure 2. 2009 Chinas date-value dataset of surface climate from weather station
1 Compute the difference of weather station information using Kriging interpolation, to form the 2009 distribution map of
mean annual temperature and mean relative humidity, as shown in Figure3.
Figure 3. 2009 distribution map of mean annual temperature and mean relative humidity a mean annual
temperature; b mean relative humidity. a
b
2015 International Workshop on Image and Data Fusion, 21 – 23 July 2015, Kona, Hawaii, USA
This contribution has been peer-reviewed. doi:10.5194isprsarchives-XL-7-W4-103-2015
104
2 Classify and vectorize the results of kriging interpolation. Then the basic regional distribution of national climate is
received, as shown in figure4.
Figure 4. Classification and vectorization of mean annual temperature and mean relative humidity a mean
annual temperature; b mean relative humidity. 3 The results of step 2 are stacked together to extract the
regionalization boundaries dominated by climate factors, as shown in Figure 5.
Figure 5. Extract the regionalization boundaries dominated by climate factors
Secondly, pre-processing of the national SRTM250m DEM data is done by segmentation, vectorization method Figure 6.
Pre-processing will produce small regions with certain similar characteristics.
Figure 6. Classification, segmentation, vectorization of national SRTM250m DEM data
2.2.3. Merge Polygons
Finally, through the overlay analysis of regionalization boundaries dominated by climate and DEM data, we merge the
small polygons, finally formed a nationwide Eco-geographical regionalization, shown in Figure 7.
Figure 7. Forming a nationwide Eco-geographical regionalization using overlay analysis
2.2.4. Result Evaluate
